Intel and Apple

by Jason Deraleau

Introducing Centrino

Intel recently backed up Apple's claim that 2003 is the year of the portable. Intel's new Centrino lineup introduces the Pentium M processor, the first Intel chip designed from the ground up for mobile systems. Traditionally, Intel's laptop chips have been "diet" versions of their desktop chips. While features like Speed Step help increase battery life, Intel (and AMD for that matter) based laptops just haven't been able to keep up with PowerPC based laptops.

PC Magazine's April 8th issue has a round up of the various new Centrino laptops coming to market. To be considered a Centrino system, a portable must pack the new Pentium M as well as a new Intel chipset that stresses wireless technology. Laptops are already being released by Acer, Dell, Gateway, and several other major players in the market. The price point on the new systems is quite a bit higher than many other portables on the market at this point (and even some Tablet PC systems).

The Pentium M is actually an interesting venture. Prior to the Pentium M, the only other major processor designed specifically for portables was Transmeta's Crusoe. If you're not familiar with the Crusoe chip, it's a hybrid processor that is half hardware and half software. The back end (the part that actually handles the calculations and processing) of the processor is the hardware portion, while the front end (the portion that handles scheduling) is based in software.

I have not yet seen a comparison that specifically makes use of Crusoe based laptops and Pentium M laptops. However, judging by the benchmarks I've seen for them on an individual basis, I'd guess that the Pentium M will out perform a Crusoe, but doesn't get as much battery life as a Crusoe system.

Why do we care?

The big concern as Mac users is how the battery life will compare to Apple's portables. It's long been considered an unwritten rule that if you wanted good battery life you should choose an Apple portable. The PowerPC processor uses far less battery power than the traditional x86 processor. While definitely not the only factor in determining battery life, most other components of Apple and x86 laptops are quite similar in design and power consumption.

Intel claims a battery life that's better than five hours. In contrast, Apple's new 12" PowerBook claims a battery life of up to five hours. I don't know about anyone else, but I've never actually gotten results even close to that. I have a 15" TiBook 800MHz and tend to get just about three hours of battery life with normal use. I'd be interested to hear how long Pentium M users are able to go without a recharge in normal use. If we go by what the manufacturers suggest at this point, it appears that the new Pentium M based machines are getting better battery life.

The Thick Plottens

Performance wise, a Pentium M machine at 1.4 GHz will best a Pentium 4 M at 2.4 GHz. Impressive, especially considering the lower power consumption. Almost makes you wonder why they don't strap these puppies into desktops so we can cut down on the number of power plants out there. My initial thoughts when I heard the name of the new processor is that it convolutes the Pentium branding. It's easy to see how people will become confused between a Pentium M and a Pentium 4 M. Two very different processors, but with names differentiated by a single character.

In addition to this, bringing in the Centrino name blurs things even further. A laptop with a Centrino label has a Pentium M processor, but apparently not all manufacturers are including the Intel Centrino chipset, which is an essential part of the platform. These half-Centrinos are shipping with the exact same label as the true Centrinos. Buyer beware!

Some additional thoughts on Intel

John Dvorak, columnist for PC Magazine, recently made some conjecture about an Apple Switch to Intel based processors. He points to various interactions between Apple and Intel over the past year or so. I'd like to just respond to this a bit, since I feel quite differently on the matter.

Apple was one of the original companies to help develop and nurture the PowerPC platform. They have spent years and millions of dollars in creating a market for the chip through their systems. They have given countless presentations on the "Megahertz Myth" to help explain how clockspeed isn't a true indicator of performance, helping to alleviate buyer concerns. Apple has a strong interest in the PowerPC processor and it is as much a part of today's Macintosh experience as a one button mouse.

A few months ago, many of the Apple rumor sites started running stories about how Apple was going to switch to using AMD processors. The rumors pointed to close ties between Apple and AMD, various common public appearances, etc. The same kind of ties that John Dvorak pointed out in his Apple Switch article. Apple keeps ties with many manufacturers on a variety of projects (e.g. Apple and AMD are both part of the HyperTransport consortium). This is normal behavior, not a sign of such a major change.

With the PowerPC 970 just around the corner, I don't see Apple switching to an x86 architecture. While it is true that many of the other components inside a Macintosh are quite similar to an x86 machine, Apple's stance is more political than technical. Apple has long kept a tight hold over the Macintosh architecture, which these days essentially encompasses the processor and chipset. Almost every other component in a Macintosh system is compliant with the same standards x86 manufacturers use for their systems.

The reason Apple keeps such a tight hold over the core of the Macintosh is to maintain the Macintosh's legendary stability. By having such specific basic hardware to work with, Mac OS X runs rock solid. I've often thought that the reason the Windows platform is so unstable is the fact that Microsoft has to write code for thousands of different processor and chipset combinations (well that and a bunch of legacy hardware and software, but I digress). Mac OS X runs on far fewer variations in hardware, helping not only with stability, but with performance (okay, in theory).

The biggest hurdle in moving your customer base to a completely different processor ISA exists in end user applications. While it is quite true that Mac OS X could easily be ported to x86 (if it hasn't been already), thousands of other Macintosh applications would have to be recompiled or rewritten to run on the different processor. Let's be realistic, not even all of the popular applications have been moved to Mac OS X yet.

Some of you might be thinking "well, why not just port the Windows stuff over?" The issue there becomes one of libraries. There is nothing like Cocoa or Carbon on Windows. Windows applications would have to be completely rewritten to run on an x86 version of Mac OS X as well. A move to x86 would require a lot of effort by both users and developers that I just don't think Apple is willing to risk forcing.

There will be a move though

Moving to the 64-bit PowerPC 970 from IBM would basically require a recompile and some small changes of Mac OS X only. The PPC 970 can handle 32-bit code (current Mac OS X) as well as 64-bit code, as long as the operating system knows how to handle both. While end user applications would all require the same recompiles and small changes, the immediate gain of having 64-bit processing in your application is minimal at best. Applications probably wouldn't move to 64-bit except in a "why not?" kind of approach.

The biggest advantage of moving to the PPC 970 doesn't come from the fact that it's 64-bit, but its higher clockspeeds. The latest reports from IBM are talking about the processor coming out with a clockspeed topping out in the 2 GHz range. This is obviously a fairly big jump from the current 1.4 GHz G4. In addition to higher clockspeeds, the PPC 970 has a lot of interesting designs to help get more work done per clock cycle than our friend the G4e. For some great information on this, I recommend John Stoke's article on IBM's PowerPC 970.

So, I do see Apple making a processor change, most likely this year even. I predict however, that it won't be to an AMD or Intel processor, but to the more logical PowerPC 970. While it won't be the fastest processor on the market either, it will definitely be a breath of fresh air for the platform. Apple is currently preparing their product line for a new processor. The G3 remains in only a single system (the iBook). Expect the PPC 970 (not sure if it will be branded the G5) in PowerMacs first, then the high end PowerBooks. The G4 will be in the "i" line of products for some time to come.

Do you think Apple will switch to x86? Does the Centrino threaten Apple's portables?


2003-03-29 05:15:38
64-bit *is* needed
The main reason to go to 64-bit platforms is that they can make use of more than 4 GB of memory. 1 GB now is pretty standard, so for this to become a basic requirement is only two revolutions of the Moore clock.
2003-03-29 08:48:18
64-bit *is* needed
With the generous ROM space and addresses reserved by the kernel, only 2gb of RAM is left for system RAM in Apple's current 32 bit PowerPC's. Going to 64 bit architecture means that Apple can cleanly access more than that without segmentation hacks. There are certainly database (including OLAP, data mining), DTP/OPI, and scientific applications that can benefit from having more than 2 gigabytes of RAM even on the relatively lightweight single and dual processor PowerPC's (as compared to, say, 8 or 32 way mid-range servers with gobs of RAM).
2003-03-29 13:32:25
Apple don't have a choice
The PowerPC 970 is just not going to be competitive with Intel, based upon projected performance from IBM and Intel.

Apple are going to have to go with Intel if they want to have decent performance. Developers will complain, but they will prefer performance.

The real question - that keeps Steve Jobs up at night - is does this increase the viability of Virtual PC?

2003-03-29 15:45:32
Apple don't have a choice
In terms of performance, the PPC 970 already has greater performance than the Intel and AMD chips. Clock speeds will continue to go up and new processors are already in the works. IBM's PPC roadmap looks pretty good.

I seriously doubt that Steve Jobs is having a hard time sleeping.

2003-03-29 18:32:05
"more political than technical"??????
You say that "Apple's stance is more political than technical," but then you proceed to list 2 points that I would say are very technical -- the stability issue, and the "having to recompile every app under the sun, and still ship PPC versions for the older computers, too" issue.

If Apple moved to an Intel processor from a PowerPC, it would, most likely, kill off Apple as a company because of these little, niggly, "political" issues.

2003-03-29 18:48:28
"more political than technical"??????
Bad ordering I suppose. The political issues had been discussed in the previous paragraph, those being the fact that Apple was one of the original PPC pushers and has spent a lot of time evangelizing the platform.
2003-03-30 00:05:34
Apple don't have a choice
"In terms of performance, the PPC 970 already has greater performance than the Intel and AMD chips."

Where are the benchmarks? It really bugs me how so many people are talking about the 970 like it's a production processor. It's not. Anything anyone says in public about 970 performance is pure hearsay and is essentially baseless.

2003-03-30 06:31:33
64-bit *is* needed
Actually, G4's (and Intel P4) have 36 bit addressing and so can access up to 64 GB of memory
2003-03-30 07:41:35
Apple don't have a choice
You want spec's, check out Daivd T. Wang's article here:

He attended IBM presentation on the PowerPC 970. Go to the second page and you will see the SPECint 2000, SPECfp2000, and the Dhrystone performance benchmarks. This chips hawls butt.

2003-03-30 10:16:46
64-bit *is* needed DATA or ADDRESS???
I thought that what we were talking about was the word width of the processor, i.e. how much data moves through the processor core in a single cycle. 64 bit processors would do twice as much data per cycle as a 32 bit processor. This is not the same as increasing the number of address bits (and therefore the amount of memory accessible).
A 32 bit application running on a 64 bit processor would still get 64 bits of data across the EXTERNAL bus at a time but each instruction and internal register would only handle 32 bits of that data at a time internally.

At least that's the way I think it works.

2003-03-31 05:30:48
There *is* something like cocoa on windows
It's called AppKit and it's been available on Windows as part of WebObjects for quite some time.

In older versions of WebObjects, AppKit was on par with YellowBox apps for Rhapsody, the predecessor to Cocoa.

So while there isn't currently the ability to run current Cocoa applications on Windows, it is certainly possible, should Apple decide to release an update for AppKit.

2003-04-01 08:59:47
There *is* something like cocoa on windows
Interesting. Never heard of it. Does this use Objective C for coding? Or is it just the Cocoa libraries? I wouldn't think Cocoa would actually be possible on Windows because there would need to be a Cocoa -> Win32 bridge. Maybe that's what you're talking about does.